Hi Geert,

> > +       dmac->dev->dma_parms = &dmac->parms;
> > +       dma_set_max_seg_size(dmac->dev, 0x01000000);
> 
> That is one too much, cfr.
> 
>     drivers/dma/sh/rcar-dmac.c:#define RCAR_DMATCR_MASK             0x00ffffff

I see. Will fix and send an updated patch. Nice to see that I was not
totally wrong but just off-by-one.

> And yes, this depends on the data length, which is not known at probe time.
> Of course, dma_set_max_seg_size(dmac->dev, RCAR_DMATCR_MASK)
> should be safe, albeit (slightly) suboptimal.

True. But still *way* better than the current default of 64K.

Thanks for the heads up!

Regards,

   Wolfram

Attachment: signature.asc
Description: PGP signature

Reply via email to